SmartQuant logo SmartQuant

SmartQuant Product

QuantMCP

The SmartQuant MCP (Model Context Protocol) server that exposes trading infrastructure, data, and logs to AI agents.

Overview

QuantMCP is the SmartQuant MCP (Model Context Protocol) server that exposes trading infrastructure, data, and logs to AI agents.

QuantMCP allows AI (such as GitHub Copilot, ChatGPT, or any MCP client) to do more than read strategy code. It can work with a live trading system: access instruments, market data, strategy logs, application state, and call framework functions exactly like a developer or operator would.

It is the bridge between the event-driven SmartQuant infrastructure and agent-based AI.

What QuantMCP does

QuantMCP provides AI with a standardized set of tools (functions) that allow it to:

  • Retrieve instruments and their parameters.
  • Read and write market data series.
  • Analyze strategy and application logs.
  • Inspect the state of running applications.
  • Control strategies through the controller.
  • Access QuantBase as a persistent event store.

AI does not “read files” and does not “look at UI”. It works through function calls like a full SmartQuant client.

Key capabilities

Access to QuantBase data

  • • Retrieve the list of available series
  • • Read Quote, Trade, and Log events
  • • Analyze strategy execution history
  • • Request specific historical segments for analysis

Access to the trading framework

Through QuantMCP, AI can access:

  • • InstrumentManager — instruments and parameters
  • • DataManager — market data and series
  • • Controller — application and strategy state

AI sees the system the same way a developer does.

Working with strategy logs

  • • Read Output events produced by Log() calls
  • • Analyze behavior in history
  • • Identify root causes of issues
  • • Suggest parameter or code changes

This enables time-travel debugging with AI.

Infrastructure monitoring

  • • See which applications are running
  • • Monitor infrastructure state
  • • Generate alerts
  • • Assist in incident investigation

How it works

There is no “AI magic” inside.

  1. QuantMCP publishes a list of functions and their descriptions.
  2. The MCP client (Copilot, ChatGPT, etc.) receives this list.
  3. The model generates text that contains a function call.
  4. The client invokes the QuantMCP function.
  5. The result is returned to the model as text.
  6. The cycle repeats.

AI interacts with the system through regular API calls expressed in text form.

Typical use cases

1) AI-assisted strategy debugging in Visual Studio

Copilot can request strategy Output and logs (via QuantMCP), analyze them, propose code changes, and repeat the loop until the desired result is achieved.

2) Trading data analysis via Copilot

Ask questions like: what instruments exist, what data series are available, what trades occurred for a given instrument, or how a strategy behaved during a specific period — without code and without UI.

3) AI-based operational monitoring

AI can analyze logs, monitor strategy state, detect anomalies, and assist operators in troubleshooting.

4) Direct connection to ChatGPT via HTTP

QuantMCP can operate not only via stdio but also via HTTP, allowing direct use from a browser or mobile app.

Role in the SmartQuant architecture

If QuantController is the brain of the system and QuantBase is the memory of the system, then QuantMCP is the AI interface to the system.

It makes the entire SmartQuant infrastructure accessible to agent-based interaction.

Integration

GitHub Copilot

SmartQuant ships with:

  • • copilot-instructions.md — API description for AI
  • • Example strategies
  • • Ready-to-use .mcp.json configuration

Once MCP is enabled in Copilot settings, AI gains access to QuantMCP tools directly inside Visual Studio.

AI strategies (AIStrategy)

The SmartQuant.AI library follows the same principles: the strategy communicates with AI via API, and AI can use QuantMCP to analyze logs and data.

Strategy → AI → QuantMCP → data/logs → AI → Strategy

What QuantMCP enables

  • • A new way of development — AI inside the strategy development loop
  • • A new way of analysis — natural-language access to data and logs
  • • A new way of monitoring — AI as an infrastructure operator
  • • Readiness for the agent-driven future — where AI becomes the primary system interface

QuantMCP is not a utility. It is a layer that makes SmartQuant ready for the era of AI agents.